qwen3-tts: switch default to -CustomVoice variant; chatterbox: fix paralinguistic tag list in README

qwen3-tts: deploy was using the -Base checkpoint, which sounds like
the right one ("supports voice cloning") but the upstream wrapper's
only synthesis path goes through generate_custom_voice. The -Base
variant doesn't expose that, so every request — including ones with
the wrapper's listed built-in voices like Ryan/Vivian — errored with
"does not support generate_custom_voice". The -CustomVoice variant
exposes both the cloning machinery and the preset voices, and is
what the wrapper actually needs.

The .env.example comments had the variant labels backward; fixed in
this commit. Live host already updated to -CustomVoice via direct
.env edit (model downloaded on container restart).

chatterbox README listed [whisper] and [breath] as supported tags —
those are in the base Chatterbox tag set but NOT in the Turbo set
that's actually loaded. Replaced with the canonical 9-tag list
verified against /api/model-info: laugh, chuckle, sigh, gasp, cough,
clear throat, sniff, groan, shush.

stacks/chatterbox/README.md

@@ -30,12 +30,16 @@ real-time voice-agent use cases.
 - **Zero-shot voice cloning from ~5 s reference** (base Chatterbox
   needs ~10 s; Turbo cuts that in half).
 - **Native paralinguistic tags inline in text** — drop these into your
-  prompt and the model honors them:
+  prompt and the model honors them. Turbo ships exactly **9** tags
+  (verified against the live model at `/api/model-info`):
   ```
-  [laugh] [cough] [sigh] [gasp] [whisper] [breath]
+  [laugh] [chuckle] [sigh] [gasp] [cough] [clear throat] [sniff] [groan] [shush]
   ```
-  Different shape from IndexTTS-2's 8-vector emotion control: cleaner
-  for "say it like this" markup directly in the prompt.
+  Note: base Chatterbox docs list `[whisper]` and `[breath]` — those
+  are NOT in the Turbo set, ignore them or you'll burn time wondering
+  why nothing happens. Different shape from IndexTTS-2's 8-vector
+  emotion control: cleaner for "say it like this" markup directly in
+  the prompt.
 - **Mandatory PerTh watermark** on outputs (Resemble policy, cannot
   be disabled). Non-issue for internal use; mention it if you ever
   ship Chatterbox-generated audio externally.

stacks/qwen3-tts/.env.example

@@ -26,15 +26,22 @@ QWEN3_TTS_BIND=0.0.0.0
 QWEN3_TTS_BACKEND=official
 
 # Model variant. Upstream publishes four checkpoints on HF:
-#   Qwen/Qwen3-TTS-12Hz-1.7B-Base        — flagship, supports voice cloning
-#   Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice — flagship, preset voices only
-#   Qwen/Qwen3-TTS-12Hz-0.6B-Base        — lightweight, supports voice cloning
-#   Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice — lightweight, preset voices only
+#   Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice — flagship, supports voice cloning + preset voices  (RECOMMENDED)
+#   Qwen/Qwen3-TTS-12Hz-1.7B-Base        — flagship, foundation only — REJECTS all synthesis through this wrapper
+#   Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice — lightweight cloning-capable variant
+#   Qwen/Qwen3-TTS-12Hz-0.6B-Base        — lightweight foundation only
 # 1.7B = ~6–8 GB VRAM bfloat16, best quality. 0.6B = ~2–3 GB.
-# `voice="clone:<name>"` only works with the -Base variants — that's
-# what the upstream wrapper enforces. Earlier versions of this file
-# had the variant labels reversed; corrected 2026-04-24.
-QWEN3_TTS_MODEL=Qwen/Qwen3-TTS-12Hz-1.7B-Base
+#
+# IMPORTANT: the wrapper's only synthesis code path goes through the
+# voice cloning machinery. The -Base variants do NOT expose
+# generate_custom_voice; they error with "does not support
+# generate_custom_voice" on every request, no matter what voice you
+# specify (built-ins like Ryan, Vivian etc. count as custom voices).
+# Use -CustomVoice. The Base variant is foundation-model territory
+# that needs a different harness to be useful.
+# (This file had the labels backwards through 2026-04-27 — fixed
+# after the 2026-04-27 voice-fleet bench surfaced the failure mode.)
+QWEN3_TTS_MODEL=Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
 
 # Warm the model on container start so the first synthesis request
 # doesn't pay the load latency. Adds ~30 s to startup. Recommended.